When compiled with clang-3.8 on ppc, compel complains:
> GEN criu/pie/parasite-native-blob.h
> parasite_native_blob: Error (compel/src/lib/handle-elf-host.c:541):
> Unsupported relocation of type 26
Look into R_PPC64_REL64/R_PPC64_ADDR64 and handle 32-bit the same way.
travis-ci: success for PPC+clang compile fixes
Cc: Laurent Dufour <ldufour@linux.vnet.ibm.com>
Signed-off-by: Kir Kolyshkin <kir@openvz.org>
Signed-off-by: Pavel Emelyanov <xemul@virtuozzo.com>
Signed-off-by: Andrei Vagin <avagin@virtuozzo.com>
In a manner similar to one of commit 9303ed3 ("Makefiles: move
-Wa,--noexecstack out of CFLAGS"), let's move this option to
linker flags (see the abovementioned commit for reasoning).
travis-ci: success for series starting with [1/3] compel: mv noexecstack from CFLAGS to LDFLAGS
Signed-off-by: Kir Kolyshkin <kir@openvz.org>
Signed-off-by: Pavel Emelyanov <xemul@virtuozzo.com>
Signed-off-by: Andrei Vagin <avagin@virtuozzo.com>
While writing this, I somehow managed to miss the check of
how many entries were saved in core image.
So it may dereference here bs.
Fixes: #228
Fixes: commit 6fde3b8c27db ("x86: restore TLS")
travis-ci: success for x86, tls: read no more than saved TLS entries
Cc: Andrei Vagin <avagin@virtuozzo.com>
Cc: Cyrill Gorcunov <gorcunov@openvz.org>
Reported-by: Andrei Vagin <avagin@virtuozzo.com>
Signed-off-by: Dmitry Safonov <dsafonov@virtuozzo.com>
Acked-by: Cyrill Gorcunov <gorcunov@openvz.org>
Signed-off-by: Pavel Emelyanov <xemul@virtuozzo.com>
Signed-off-by: Andrei Vagin <avagin@virtuozzo.com>
The problem is, -Wa is a flag for assembler, but CFLAGS are also used
to generate dependencies, and clang complains loudly when it is used
for deps:
> > DEP compel/arch/x86/plugins/std/syscalls-64.d
> > clang-3.8: error: argument unused during compilation:
> > '-Wa,--noexecstack'
This patch moved the noexecflag from assembler to linker. I am not
100% sure but the end result seems to be the same.
This fixes dependency generation when using clang instead of gcc.
I surely have done my research before proposing this change, and
I have tested this change as good as I could.
Sorry, I should have provided more background in the commit message.
Here it goes.
There are a few ways to have non-executable stack:
1. mark the assembler source file (.S) with .section
.note.GNU-stack,"",%progbits
2. pass the -Wa,--noexecstack to compiler
3. pass the -z execstack to linker
All three ways are fine, let's see them in greater details.
Some people say (1) is the best way, but we have way too many
.S files now (23 of them, to be exact). Anyway, I can certainly do it
this way if you like, just let me know. It would look like this:
--- a/compel/arch/aarch64/plugins/std/syscalls/syscall-aux.S
+++ b/compel/arch/aarch64/plugins/std/syscalls/syscall-aux.S
@@ -3,6 +3,8 @@
* that are not implemented in the AArch64 Linux kernel
*/
+.section .note.GNU-stack,"",%progbits
+
ENTRY(sys_open)
mov x3, x2
mov x2, x1
Way (2) is what is currently used. Unfortunately it breaks dependency
generation with clang. One way to fix it would be to filter-out the bad
flag when we're generating deps. I tried experimenting with
$(filter-out) function in Makefiles today but it's complicated and I failed
to make it work.
Way (3) is what this commit offers. It seem to work fine while being
the least intrusive.
Signed-off-by: Kir Kolyshkin <kir@openvz.org>
Reviewed-by: Cyrill Gorcunov <gorcunov@openvz.org>
Reviewed-by: Dmitry Safonov <dsafonov@virtuozzo.com>
Signed-off-by: Pavel Emelyanov <xemul@virtuozzo.com>
Signed-off-by: Andrei Vagin <avagin@virtuozzo.com>
Got the following from clang:
> CC criu/parasite-syscall.o
> criu/parasite-syscall.c:1494:6: error: variable 'parasite_size' is used
> uninitialized whenever 'if' condition is false
> [-Werror,-Wsometimes-uninitialized]
> if (seized_native(ctl))
> ^~~~~~~~~~~~~~~~~~
> criu/parasite-syscall.c:1503:2: note: uninitialized use occurs here
> parasite_size += ctl->args_size;
> ^~~~~~~~~~~~~
> criu/parasite-syscall.c:1494:2: note: remove the 'if' if its condition
> is always true
> if (seized_native(ctl))
> ^~~~~~~~~~~~~~~~~~~~~~~
I look at the code and I'm still not quite sure whether parasite_size
can end up being uninitialized. Looks like for CONFIG_X86_64 it can.
So, let's initialize it.
Signed-off-by: Kir Kolyshkin <kir@openvz.org>
Signed-off-by: Pavel Emelyanov <xemul@virtuozzo.com>
Signed-off-by: Andrei Vagin <avagin@virtuozzo.com>
This patch fix a build break on ppc64le introduced by changes in the
compel's build option.
I factorized the common CFLAGS to make easier identifying particular
flags required by some architectures.
Fixes: fef01d79f7ee ("compel: cli -- Add more flags into cflags report")
travis-ci: success for compel: fix build break on ppc64le
Cc: Cyrill Gorcunov <gorcunov@openvz.org>
Signed-off-by: Laurent Dufour <ldufour@linux.vnet.ibm.com>
Acked-by: Cyrill Gorcunov <gorcunov@openvz.org>
Signed-off-by: Pavel Emelyanov <xemul@virtuozzo.com>
Signed-off-by: Andrei Vagin <avagin@virtuozzo.com>
When compel build objects it provides ".compel.init" and
".compel.exit" setions for plugins init/exit routines
and when we link PIE code in criu we simply zap them
out but code still reference to them inside std
text (even if we don't call them) so dont drop them.
Eventually compel will be loading plugins by self but
for now it's good for testing purpose to refernce
the plugin inside criu code.
Signed-off-by: Cyrill Gorcunov <gorcunov@openvz.org>
Reviewed-by: Dmitry Safonov <dsafonov@virtuozzo.com>
Signed-off-by: Pavel Emelyanov <xemul@virtuozzo.com>
Signed-off-by: Andrei Vagin <avagin@virtuozzo.com>
@std plugin known to fail when trying to execute,
investigating now. Meanwhile to not block tests
lets use syscalls object file.
Signed-off-by: Cyrill Gorcunov <gorcunov@openvz.org>
Signed-off-by: Pavel Emelyanov <xemul@virtuozzo.com>
Signed-off-by: Andrei Vagin <avagin@virtuozzo.com>
And use it in CRIU directly instead:
- move syscalls into compel/arch/ARCH/plugins/std/syscalls
- drop old symlinks
- no build for 32bit on x86 as expected
- use std.built-in.o inside criu directly (compel_main stub)
- drop syscalls on x86 criu directory, I copied them already
in first compel commist, so we can't move them now, but
delete in place
Signed-off-by: Cyrill Gorcunov <gorcunov@openvz.org>
Signed-off-by: Pavel Emelyanov <xemul@virtuozzo.com>
Signed-off-by: Andrei Vagin <avagin@virtuozzo.com>
Real syscalls generation is inside criu for a while
but will be moved out in the next patch.
Signed-off-by: Cyrill Gorcunov <gorcunov@openvz.org>
Signed-off-by: Pavel Emelyanov <xemul@virtuozzo.com>
Signed-off-by: Andrei Vagin <avagin@virtuozzo.com>
Both std and criu will use it for syscalls sake.
Note I've to disable x86 compat mode for a while:
we have to provide native types there thus will
back once everything else is complete.
Signed-off-by: Cyrill Gorcunov <gorcunov@openvz.org>
Signed-off-by: Pavel Emelyanov <xemul@virtuozzo.com>
Signed-off-by: Andrei Vagin <avagin@virtuozzo.com>
- linkage.h points to CRIU's asm specific
- prologue.h is just a link to x86, we will need to
fix it when start using std with other plugins
Signed-off-by: Cyrill Gorcunov <gorcunov@openvz.org>
Signed-off-by: Pavel Emelyanov <xemul@virtuozzo.com>
Signed-off-by: Andrei Vagin <avagin@virtuozzo.com>
As it's a make/build option, the set of files to compile depends
on it. We don't do feature tests on mrproper/clean (no point in it).
Let's define CONFIG_COMPAT for those targets, so compiled files
will be properly cleaned.
Reported-by: Adrian Reber <adrian@lisas.de>
Cc: Cyrill Gorcunov <gorcunov@openvz.org>
Signed-off-by: Dmitry Safonov <dsafonov@virtuozzo.com>
Acked-by: Cyrill Gorcunov <gorcunov@openvz.org>
Signed-off-by: Pavel Emelyanov <xemul@virtuozzo.com>
Signed-off-by: Andrei Vagin <avagin@virtuozzo.com>
As -lsoccr now is in $(LIBS) in Makefile.packages, it should be
build before criu objects. Otherwise, we try to compile
feature-tests with $(LIBS) and the succesful features fails with:
<stdin>:1:25: warning: extra tokens at end of #include directive
/usr/bin/ld: cannot find -lsoccr
collect2: error: ld returned 1 exit status
Which lead to later problems:
criu/sk-tcp.c:50:8: error: redefinition of 'struct tcp_repair_opt'
struct tcp_repair_opt {
^
In file included from criu/sk-tcp.c:1:0:
/usr/include/netinet/tcp.h:259:8: note: originally defined here
struct tcp_repair_opt
Because of fails in really-sucessful feture-tests.
P.S.
Maybe we should unbound feature-tests compilation from $(LIBS),
specified in Makefile.packages - that looks right after all.
Cc: Cyrill Gorcunov <gorcunov@openvz.org>
Cc: Pavel Emelyanov <xemul@virtuozzo.com>
Signed-off-by: Dmitry Safonov <dsafonov@virtuozzo.com>
Signed-off-by: Pavel Emelyanov <xemul@virtuozzo.com>
Signed-off-by: Andrei Vagin <avagin@virtuozzo.com>
compel-host is needed to build criu itself and
run during piegen blob generation.
Signed-off-by: Cyrill Gorcunov <gorcunov@openvz.org>
Signed-off-by: Pavel Emelyanov <xemul@virtuozzo.com>
Signed-off-by: Andrei Vagin <avagin@virtuozzo.com>
The default compel binary is the one compiled for target architecture.
The compel-host one is compel compiled for current arch that is needed
to build something for target arch.
We could use default compel if target arch == current arch, but this
makes things more complex.
Signed-off-by: Cyrill Gorcunov <gorcunov@openvz.org>
Signed-off-by: Pavel Emelyanov <xemul@virtuozzo.com>
Signed-off-by: Andrei Vagin <avagin@virtuozzo.com>
The shmem pluging allows creation of shared memory segment between
parasite code and the caller.
Signed-off-by: Dmitry Safonov <dsafonov@virtuozzo.com>
Signed-off-by: Pavel Emelyanov <xemul@virtuozzo.com>
Signed-off-by: Andrei Vagin <avagin@virtuozzo.com>
The prologue includes routines needed for parasite blob to work
and is always included with the std plugin.
Signed-off-by: Dmitry Safonov <dsafonov@virtuozzo.com>
Signed-off-by: Pavel Emelyanov <xemul@virtuozzo.com>
Signed-off-by: Andrei Vagin <avagin@virtuozzo.com>
Delete plugins/include/asm/std directory - let it be without plugin name.
Make symlinks to reuse criu's files, except those, which will
be deleted after libcompel from criu (like syscalls).
Signed-off-by: Dmitry Safonov <dsafonov@virtuozzo.com>
Signed-off-by: Pavel Emelyanov <xemul@virtuozzo.com>
Signed-off-by: Andrei Vagin <avagin@virtuozzo.com>
The plugin provides basic features as string copying, syscalls, printing.
Not used on its own by now but will be shipping by default with other
plugins.
With great help from Dmitry Safonov.
Signed-off-by: Cyrill Gorcunov <gorcunov@openvz.org>
Signed-off-by: Pavel Emelyanov <xemul@virtuozzo.com>
Signed-off-by: Andrei Vagin <avagin@virtuozzo.com>
As we meet this before, I think, we should abort generation here.
No point for making PIE with external symbols for us (at least, yet),
so let's break build than find the problem at runtime.
Cc: Cyrill Gorcunov <gorcunov@openvz.org>
Signed-off-by: Dmitry Safonov <dsafonov@virtuozzo.com>
Acked-by: Cyrill Gorcunov <gorcunov@openvz.org>
Signed-off-by: Pavel Emelyanov <xemul@virtuozzo.com>
Signed-off-by: Andrei Vagin <avagin@virtuozzo.com>
It's used in restorer to map compatible vDSO blob.
When !CONFIG_COMPAT it's not needed and even more:
vdso_fill_symtable_compat is macro and is undefined in that
case. It will never shoot, as compat_vdso is always false
for that case, but get rid of it.
I also need to leave CONFIG_X86_64 here, as arch_prctl()
is not defined for x86_32.
Cc: Cyrill Gorcunov <gorcunov@openvz.org>
Signed-off-by: Dmitry Safonov <dsafonov@virtuozzo.com>
Signed-off-by: Pavel Emelyanov <xemul@virtuozzo.com>
Signed-off-by: Andrei Vagin <avagin@virtuozzo.com>
The other way to solve this is to compile shared between
CRIU and PIEs objects twice: without stack-protection and
with environment choice. But I don't think it worth it.
Cc: Cyrill Gorcunov <gorcunov@openvz.org>
Signed-off-by: Dmitry Safonov <dsafonov@virtuozzo.com>
Signed-off-by: Pavel Emelyanov <xemul@virtuozzo.com>
Signed-off-by: Andrei Vagin <avagin@virtuozzo.com>
I think, we could transpher also `-m elf_i386' and `-m32' to
compel output - but only in case if we will not support
native x86_32 in compel (as they are wrong for native build).
Or introduce something like compat_ia32 in compel option, not sure.
Left in makefile for a while.
Also it would be good to remove $(LD_R) in makefile and use
compel output for ldflags, but that will be valid only when compel
will support arm arch.
Fixes (with the following patches):
GEN criu/pie/parasite-compat-blob.h
parasite_compat_blob: Error (compel/handle-elf-32.c:322): Unexpected undefined symbol: `__stack_chk_fail'. External symbol in PIE?
Cc: Cyrill Gorcunov <gorcunov@openvz.org>
Signed-off-by: Dmitry Safonov <dsafonov@virtuozzo.com>
Signed-off-by: Pavel Emelyanov <xemul@virtuozzo.com>
Signed-off-by: Andrei Vagin <avagin@virtuozzo.com>
